KR102426291B1 - Prediction method and system of hypertension using machine learning, computer program for the same - Google Patents

Abstract

The present invention relates to a hypertension prediction method and system performed in a hypertension prediction system using machine learning, and receives a patient's ballistic signal from a ballistic sensor installed in a bed, extracts biometric information, and performs a plurality of machine learning (machine learning) ) relates to a hypertension prediction method and system performed in a hypertension prediction system using machine learning that determines a high blood pressure alert level from high blood pressure prediction probabilities calculated by input into an algorithm and provides it to a patient.

Description

Hypertension prediction method and system performed in a hypertension prediction system using machine learning, and a computer program for performing the same

The present invention relates to a hypertension prediction method and system performed in a hypertension prediction system using machine learning, and receives a patient's ballistic signal from a ballistic sensor installed in a bed, extracts biometric information, and performs a plurality of machine learning (machine learning) ) relates to a hypertension prediction method and system performed in a hypertension prediction system using machine learning that determines a high blood pressure alert level from high blood pressure prediction probabilities calculated by input into an algorithm and provides it to a patient.

Blood pressure is the pressure exerted on the walls of blood vessels by blood flowing through them, and is one of the important vital signs. Since high blood pressure in which blood pressure is abnormally high or hypotension that is abnormally low is not only a disease that needs to be managed by itself, but also acts as a cause or risk factor for various other diseases, accurate monitoring of blood pressure is very important for maintaining health. can do.

Meanwhile, the invention of chronic diseases resulting from the improvement of living standards and westernized eating habits is increasing worldwide, and among them, the number of patients with high blood pressure is steadily increasing. High blood pressure is caused by various environmental factors such as bad eating habits, family history, drinking, smoking, lack of exercise, and stress, and can be broadly divided into two types.

One is 'essential hypertension', which occurs without a causative disease, and the other is 'secondary hypertension', which occurs secondary to kidney disease, vascular abnormalities, or endocrine diseases. The incidence of secondary hypertension among all hypertensive patients is estimated to be less than 5%, and among them, hypertension due to kidney disease accounts for a high frequency. Conversely, if essential hypertension persists, the probability of developing kidney disease increases. Therefore, in order to prevent the onset of other diseases and to treat hypertension, it is necessary to continuously determine hypertension.

For the determination of hypertension, a blood pressure level is usually measured at a hospital or home using a blood pressure measuring device, and hypertension is determined according to the measured blood pressure level. However, if there is no blood pressure measuring device, the above-described determination method cannot be used, and there is a problem in that it is difficult to measure blood pressure immediately in an emergency situation.

On the other hand, a technique for checking blood pressure by acquiring biometric information in an unconstrained method rather than a structure worn on a patient's body like a blood pressure measuring device is disclosed. However, it was not possible to objectively predict the onset of hypertension because it was simply to acquire and confirm biometric information.

Unexamined Patent Publication No. 10-2005-0057593 (June 16, 2005, Hypertension Risk Diagnosis Method) Unexamined Patent Publication No. 10-2014-0039408 (April 2014.04.02. Apparatus and method for generating high blood pressure judgment model using body type information, device and method for determining hypertension using high blood pressure judgment model)

The present invention has been devised to solve the above problems, by attaching a ballistic sensor to a patient's bed or mattress to acquire biometric information in an unconstrained way without measuring blood pressure through the patient's body, and It is to provide a hypertension prediction method and system performed in a hypertension prediction system using machine learning that can not only check a patient's blood pressure level by using it, but also predict the occurrence of hypertension.

In order to solve the above problems, the hypertension prediction method performed in the hypertension prediction system using machine learning of the present invention is a method for predicting hypertension using a patient's ballistic signal obtained from a ballistic sensor installed in a bed or mattress. The method comprising: receiving, by a computer, a patient's age and gender, and a ballistic ballistic signal for predicting hypertension; Each average is calculated by extracting biometric information, such as heart rate, respiration rate, relative stroke volume, and heart rate variability, from the trajectory signal, and according to age and gender calculating a patient's hypertension prediction probability by input to a plurality of pre-trained machine learning algorithms; and outputting a high blood pressure warning level by determining whether a hypertension prediction probability having the highest value among the calculated hypertension prediction probabilities falls within a set threshold range.

Here, calculating the hypertension prediction probability; It further includes; removing noise from the previously collected trajectory signal.

In this case, the step of removing the noise; scaling the size of the trajectory signal; extracting a trajectory signal corresponding to a preset root mean square maximum value and a preset root mean square minimum value range from the scaled trajectory signal; and extracting a trajectory signal corresponding to a preset minimum amplitude value and a maximum amplitude value range from the extracted trajectory signal.

On the other hand, the high blood pressure alert level is classified into relief, interest, and attention, and when the hypertension prediction probability is 0% to 10%, "safe", when the hypertension prediction probability is 11% to 30%, "interest", the hypertension If the prediction probability is more than 31%, it is judged as "caution" and output.

And the hypertension prediction system using machine learning of the present invention is a bed; a trajectory sensor installed on the bed to acquire a trajectory signal of the patient; a server storing the computer program of claim 8 for predicting hypertension by receiving the ballistic signal acquired from the ballistic sensor; and a smart device in which an application for receiving and outputting hypertension prediction results from the server is installed.

According to the hypertension prediction method and system performed in the hypertension prediction system using machine learning of the present invention having the above configuration, the patient's ballistic signal is received from the ballistic sensor installed in the bed, and the blood pressure value is checked in real time as well as hypertension. Since the onset of the disease can be predicted, the patient does not need to wear a measuring device. In particular, the accuracy is improved because the probability of hypertension extracted through multiple machine learning (machine learning) algorithms is compared with each other to determine whether or not hypertension occurs. It has the advantage of being high.

In addition, the patient can check the blood pressure level just by lying or sitting in bed, thereby predicting the onset of hypertension, thereby diagnosing hypertension early and performing health management of the patient.

1 is a schematic configuration diagram showing a hypertension prediction system using a ballistic signal according to an embodiment of the present invention;
2 is a flowchart of a method for predicting hypertension using a ballistic signal according to an embodiment of the present invention;
3 is a flowchart of a method for removing noise of a deep trajectory signal according to an embodiment of the present invention;
4 is a graph showing a trajectory signal from which noise has been removed according to an embodiment of the present invention;
5 is a ROC-Curve graph showing a performance index of a learned artificial neural network according to an embodiment of the present invention;
6 is a ROC-Curve graph showing a performance index of a learned logistic regression analysis according to an embodiment of the present invention.

Hereinafter, a hypertension prediction method and system performed in a hypertension prediction system using machine learning according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

As shown in the figure, the hypertension prediction method and system performed in the high blood pressure prediction system using machine learning according to the present invention receive the age, sex, and ballistic signals of the patient from the ballistic sensor installed in the bed and extract biometric information It is a hypertension prediction method and system that can check whether hypertension occurs from the high blood pressure probability output by inputting biometric information into a plurality of machine learning (machine learning) algorithms.

In addition, the method for predicting hypertension according to an embodiment of the present invention is performed by a computer, and the computer stores a computer program for executing the method for predicting hypertension by operating the computer.

In addition, the computer refers to a computing device in a broad sense including a smart device such as a smart phone or a tablet PC, as well as a general personal computer.

In addition, the computer program may be stored in a separate recording medium or stored in a cloud server connected to the Internet, and the recording medium is specially designed and configured for the present invention, or a person having ordinary knowledge in the field of computer software It may be known to and available for use.

For example, the recording medium may include a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, an optical recording medium such as a CD or DVD, a magneto-optical recording medium capable of both magnetic and optical recording, a ROM, a RAM, and a flash memory. and the like, alone or in combination, may be a hardware device specially configured to store and execute program instructions.

In addition, the computer program may be a computer program composed of program instructions, a local data file, a local data structure, etc. alone or in combination, and may be executed by a computer using an interpreter as well as machine code such as generated by a compiler. It may be a computer program written in a high-level language code that can be

1 shows a hypertension prediction system according to an embodiment of the present invention. Referring to FIG. 1 , the hypertension prediction system of the present invention includes a bed 10, a ballistic sensor 20, a server 30, and a smart device ( 40) consists of

The bed 10 is composed of a bed frame 11 and a mattress 12, and provides a space for the patient to lie down.

The ballistic sensor 20 may be installed in the bed 10 , measure a patient's ballistic signal, and transmit it to the server 30 .

In addition, the trajectory sensor 20 may be installed on the side or lower surface of the bed frame 11 , and may be used while being attached to the mattress 12 in addition to the bed frame 11 .

On the other hand, the trajectory signal is an electrical recording of the body movement that occurs according to the physical movement of the heart, and represents the amount of movement of blood ejected by the heart. Measure the generated micro-vibration.

The server 30 is installed with the computer program, and receives a ballistic signal from the ballistic sensor 20 through a communication network to predict whether hypertension develops.

The smart device 40 receives the predicted occurrence of hypertension and biometric information from the server 30 through a communication network, and outputs it so that the patient can monitor it.

In addition, the smart device 40 may be installed with an application capable of monitoring the occurrence of hypertension by displaying and monitoring biometric information.

Hereinafter, the hypertension prediction method of the present invention will be described in detail with reference to FIG. 2 .

2 is a flowchart of a method for predicting hypertension according to an embodiment of the present invention. Referring to FIG. 2 , the method for predicting hypertension of the present invention largely includes a step (S1000) of receiving a patient's age, sex, and ballistic signal, the hypertension probability and calculating (S2000) and determining and outputting a high blood pressure alert level (S3000).

First, the hypertension prediction method of the present invention receives a patient's ballistic signal from the ballistic sensor 20 and also receives the patient's age and gender information (S1000).

The ballistic signal may be a ballistic signal in a state in which the patient is awake or a ballistic signal obtained while the patient is sleeping.

Next, the noise of the input trajectory signal is removed (S2000).

On the other hand, the input trajectory signal is affected by the bed's spring constant and damping factor when the patient lies or moves on the bed, and there is a problem of bringing the signal amplitude modulation and damping time constant. should be done

3 is a flowchart illustrating a method for removing noise from a deep trajectory signal according to an embodiment of the present invention, and FIG. 4 is a graph showing a trajectory signal from which noise has been removed according to an embodiment of the present invention.

Referring to FIGS. 3 to 4 , in order to remove the noise of the deep trajectory signal, first, the signal magnitude of the input deep trajectory signal is scaled ( S2100 ).

In this case, it is preferable to adjust the scaling of the signal size to be "10um/s ² ", thereby preventing an overflow of appropriate resolution and numbers.

Next, a sea trajectory signal corresponding to a preset minimum root mean square (RMS) value and a maximum root mean square (RMS) value is extracted from the scaled sea trajectory signal ( S2200 ).

Next, by extracting a signal corresponding to a preset minimum amplitude value and a preset maximum amplitude value from the extracted sea trajectory signal, noise removal of the finally input trajectory signal is completed ( S2300 ).

On the other hand, the preset minimum root mean square (RMS) value, maximum root mean square (RMS) value, minimum amplitude value, and maximum amplitude value are average values extracted from the trajectory signals of patients lying or sitting in bed, the patient or around the bed. It can be set in various ways according to the environment of

Next, the high blood pressure prediction probability is output by inputting the noise-removed trajectory signal to the machine learning algorithm (S3000).

In detail, first, biometric information is extracted from the trajectory signal from which the noise has been removed (S3100).

Here, the biometric information includes a heart rate, a respiration rate, a relative stroke volume, and a heart rate variability for a preset time, and includes various methods can be extracted.

Next, an average value of the extracted biometric information is calculated (S3200). At this time, it is preferable to calculate each according to age and gender.

Next, the average values of the calculated biometric information are input to a plurality of machine learning algorithms learned in advance according to age and gender, and the hypertension prediction probability is calculated and output (S3000).

The machine learning algorithm is an algorithm that allows a computer to learn by itself through input data, and in the present invention, a high blood pressure probability is calculated using a machine learning algorithm that can classify data into one of two types.

Specifically, the machine learning algorithm may use a classification technique algorithm such as regression analysis, logistic regression analysis, decision tree, Bayesian classification, artificial neural network, SVM (Suppoter Vector Machine), K-nearest neighbor, and the like, and the present invention uses an artificial neural network with the best classification performance and a logistic regression analysis algorithm.

In addition, in the present invention, 20,000 biometric information (heart rate, heart rate variability) data sets for learning along with age and gender were collected from hypertensive patients and normal people, respectively, and the artificial neural network and logistic regression analysis algorithm were respectively learned.

In addition, 20,000 sets of predictive biometric information (heart rate, heart rate variability) data sets were separately collected from hypertensive patients and normal people along with age and gender, and entered into the learned artificial neural network and logistic regression analysis algorithm to evaluate classification performance. Through this, parameters that can have high classification performance were optimized.

In addition, the artificial neural network was trained 150 times in four layer layers having 64 neurons in each layer.

5 is a ROC-Curve graph showing a performance indicator of a trained artificial neural network according to an embodiment of the present invention, and FIG. 6 is a ROC-Curve graph showing a performance indicator of a learned logistic regression analysis according to an embodiment of the present invention. It is a graph.

5 to 6 , the ROC-Curve graph is a graph showing better performance as the area increases, and the trained artificial neural network showed 93% prediction accuracy and 016% error, and the learned logistic regression analysis It can be seen that both machine learning algorithms show excellent performance with a prediction accuracy of 96%.

Therefore, the heart rate and heart rate variability extracted from the heart ballistic signal can be considered as discriminatory data for judging hypertension.

Meanwhile, the artificial neural network and the logistic regression analysis algorithm finally output each probability corresponding to hypertension, and in the present invention, the output probability is defined as the hypertension probability.

Next, after outputting the hypertension probability through the plurality of machine learning algorithms, the hypertension alert level is determined from the hypertension probability ( S4000 ).

In order to determine the high blood pressure alert level, first, the highest hypertension prediction probability is extracted by comparing the output hypertension prediction probabilities ( S4100 ).

Here, the reason for extracting the highest hypertension prediction probability is to minimize the error in judging hypertension. Rather than arbitrarily setting an error range, select the largest value among the hypertension prediction probabilities obtained through analysis. By judging high blood pressure, this is to perform high blood pressure prediction with high reliability and accuracy.

Next, it is determined whether the extracted hypertension probability falls within a preset threshold range to classify the hypertension alert level (S4200)

The high blood pressure alert level may be classified as "safe", "interest", and "caution", and when the hypertension probability is 0% to 10%, it is determined as "safe", and the hypertension probability is 11% to 30% In the case of "interest", when the hypertension probability is 31% or more, it is determined and output as "caution".

Through these steps, it is possible to diagnose whether a patient develops hypertension, and the output result can be transmitted to a smart device for monitoring.

Therefore, since the hypertension prediction method and system performed in the hypertension prediction system using machine learning of the present invention can predict the patient's hypertension by receiving the ballistic signal from the ballistic sensor installed in the bed, the patient is measured by wearing a measuring device. There is no need to do this, and it is possible to predict whether the patient has high blood pressure just by lying or sitting in bed, which has the advantage of diagnosing high blood pressure at an early stage and performing health management of the patient.

10: Bed 20: Deep ballistic sensor
30: server 40: smart device

Claims (9)

In the method of predicting hypertension by using a patient's ballistic signal obtained from a ballistic sensor installed on a bed or mattress,
receiving, by the computer, the age and gender of the patient whose hypertension is to be predicted, and the ballistic signal;
Each average is calculated by extracting biometric information, such as heart rate, respiration rate, relative stroke volume, and heart rate variability, from the trajectory signal, and according to age and gender calculating a patient's hypertension prediction probability by input to a plurality of pre-trained machine learning algorithms; and
outputting a high blood pressure warning level by determining whether a hypertension prediction probability having a highest value among the calculated hypertension prediction probabilities falls within a set threshold range;
Hypertension prediction method performed in a hypertension prediction system using machine learning, characterized in that it comprises a.
According to claim 1,
calculating the hypertension prediction probability; Before,
The hypertension prediction method performed in the hypertension prediction system using machine learning, characterized in that it further comprises; removing the noise of the collected trajectory signal.
3. The method of claim 2,
removing the noise;
scaling the magnitude of the trajectory signal;
extracting a trajectory signal corresponding to a preset root mean square maximum value and a preset root mean square minimum value range from the scaled trajectory signal; and
A method for predicting hypertension performed in a hypertension prediction system using machine learning, comprising: extracting a ballistic signal corresponding to a preset minimum amplitude value and a maximum amplitude value range from the extracted ballistic signal.
delete delete According to claim 1,
The high blood pressure alert level is classified as reassurance, concern, attention,
When the hypertension prediction probability is 0% to 10%, it is judged as “safe”, when the hypertension prediction probability is 11% to 30%, “interested”, and when the hypertension prediction probability is 31% or more, it is judged and outputted as “caution”. Hypertension prediction method performed in a hypertension prediction system using machine learning, characterized by
According to claim 1,
The plurality of machine learning algorithms are an artificial neural network (ANN) and a logistic regression analysis algorithm. A hypertension prediction method performed in a hypertension prediction system using machine learning.
A computer program stored in a recording medium to perform the hypertension prediction method using the ballistic signal of any one of claims 1, 2, 3, 6, and 7.
bed;
a trajectory sensor installed on the bed to acquire a trajectory signal of the patient;
a server storing the computer program of claim 8 for predicting hypertension by receiving the ballistic signal acquired from the ballistic sensor; and
Hypertension prediction system using machine learning, characterized in that it includes a smart device installed with an application that receives and outputs the hypertension prediction result from the server.

Images